We are thrilled to announce the winners of the Hermes Agent Challenge! Over the past few weeks, the DEV community dove headfirst into one of the most exciting open-source agentic systems on the scene, and the submissions absolutely delivered.
The range of what people built and wrote left us genuinely impressed — Hermes' memory architecture got a thorough cognitive-science breakdown, the cost curve of a self-improving agent got stress-tested in production, someone built a full immune system for stress-testing AI agent safety, and a Linux desktop HAL was shipped entirely by AI agents. These are just a few of the standout submissions that made judging genuinely hard.
Now, without further ado, let's celebrate our winners! 🎉
From an offline-first emergency response platform to a 25-agent legislative simulation, the Build submissions pushed Hermes into genuinely ambitious territory. What stood out across the board was how purposefully each project used Hermes' agentic capabilities — not as a wrapper around a chat model, but as the actual load-bearing infrastructure. Parallel task delegation, persistent memory, scheduled autonomy, and model-agnostic routing all showed up in real, working builds. The Write submissions gave us useful breakdowns and genuine perspective. The strongest pieces ranged from practical setup guides to sharp critical arguments — about the self-trust issue baked into Hermes' skill architecture, what persistent memory actually changes about the developer experience, and when local agents earn their complexity. Whether you're just getting started with Hermes or already running it in production, there's something here for you to sink your teeth into.
All eight winners will each receive:
All participants with a valid submission will receive a completion badge.
Check out the June Solstice Game Jam happening now:
Thank you once more to everyone who participated.
Happy coding! 💙